The Agent Skills Directory

[DATA_EXFILTRATION]: The skill packages sensitive local information, including full chat transcripts, git snapshots, and "hot context" (current project state). This data is stored in the user's home directory at ~/.claude/handoffs and is intended for transfer to another agent or model session.
[COMMAND_EXECUTION]: The skill triggers the execution of a Node.js script (scripts/claude-to-codex.mjs) to perform transcript resolution, file system writes, and session dispatching. The use of the direct Node CLI for advanced options also involves shell execution.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by ingesting and bridging untrusted data (transcripts and git state) between agents.
Ingestion points: The skill reads the active transcript and current git state from the local environment.
Boundary markers: While the skill mentions using a "bounded pointer prompt" rather than a full paste, it does not specify explicit delimiters or "ignore previous instructions" guards for the content being handed off.
Capability inventory: The skill possesses the capability to write to the file system and execute shell commands via Node.js.
Sanitization: There is no description of sanitization, filtering, or validation of the transcript content before it is processed by the target model.

handoff